Remote sensing assessment of urban vegetation's dust retention for mitigating atmospheric particulate pollution

Abstract

Urbanization has significantly increased particulate emissions, while vegetation plays a key role in capturing and retaining atmospheric particles, thereby improving air quality. However, the large-scale impact of urban vegetation on dust retention remains underexplored. This study integrates Sentinel-2 data with vegetation spectral data to extend the analysis from individual points to larger urban areas. Five urban vegetation species from Xuhui and Minhang districts in Shanghai were selected for dust retention and canopy reflectance data collection. Ground-based spectral measurements were converted into Sentinel-2 spectra using a spectral response function. Ten vegetation indices (VIs), including SR, NDVI, SIPI, and ARVI, were evaluated for their correlation with dust retention, and four machine learning algorithms were compared. The optimal algorithm was selected for modeling the spatial distribution of vegetation dust retention. The results indicated that: (1) NDVI, ARVI, SIPI, and SR were sensitive to dust retention, with correlation coefficients of −0.78, −0.78, −0.77, and − 0.73, respectively; (2) Random Forest outperformed the other algorithms in estimating regional dust retention, with an R² of 0.65, surpassing polynomial regression, stochastic gradient descent, and support vector machines. High dust retention areas were associated with continuous vegetation cover and urban greening. These findings provide valuable insights for urban green space planning and offer a scalable method for regional dust retention estimation.